A new bi-level quantizing method for document images

An adaptive bilevel quantizing technique for images containing text and continuous-tone areas in the same page is proposed. At first, the original document image is divided into M*M pixel blocks and whether a block is a continuous-tone area or not is determined by using two features, the density of black pixels in the block and the differential value to neighbors. Then, continuous-tone areas are made binary by using the multistep dither matrix. Finally, the remaining area, which includes text and figures, is quantized to bilevel.<
>